Rio's November 2023 heatwave is now estimated at 1,392 excess deaths

Brazil's November 2023 heat event was extreme enough to be visible from outside the country — temperatures in Rio de Janeiro reached 42°C [s1]. What it cost has taken until now to quantify. A study published in the Pan American Journal of Public Health puts the excess at 1,392 deaths in Rio de Janeiro state over a very short period [s1][s2].

The more useful part of the paper is not the total but the pattern, which departs from what heat-mortality surveillance is generally set up to detect.

How the event was defined

The authors characterised the heatwave using the excess heat factor, an index that combines how hot a three-day period is relative to recent conditions and relative to the local climatology [s2]. By that index, every region of Rio de Janeiro state experienced an excess of heat classified as low intensity, severe or extreme between 11 and 17 November 2023, and on 16 November the event was severe or extreme across the whole state [s2].

Deaths from all causes in the state peaked on 18 and 19 November, two days after the index peaked [s2]. The authors compare that peak to the historical series from 2012 to 2023 and describe it as comparable to death rates seen during the Covid-19 pandemic in 2020–21 [s2]. They note that no other event could explain the pattern [s2], and report that São Paulo and Minas Gerais showed mortality peaks two to three days after the heatwave, with full results to be published separately [s2].

The two-day lag is characteristic of heat mortality and is one reason official heat-death counts undercount: the deaths do not occur while the news is covering the heat.

The size of it

Observed deaths ran 1.23 times expected across the state during the heatwave, an increment of 23.2% over the expected mortality rate, corresponding to 1,392 additional deaths against historical records [s2].

Every one of the nine macroregions showed an observed-to-expected ratio above 1 [s2]. Excess mortality was highest in the Northwest region (74.04%), then Centro Sul (41.27%), Metropolitana II (27.51%) and Metropolitana I (22.73%) [s2].

A separate model relating mortality to a composite heat indicator found the Northwest region's mortality rate elevated by 84% (rate ratio 1.84, 95% CI 1.52 to 2.19) between 8 and 22 November, followed by Centro Sul (1.27, 1.11 to 1.45), Metropolitana II (1.22, 1.12 to 1.33) and Médio Paraíba (1.18, 1.08 to 1.28) [s2]. Metropolitana I was borderline (1.13, 1.00 to 1.27, p=0.051) and the remaining regions were not statistically significant [s2].

The causes are the finding

Heat mortality is usually expected to surface as cardiovascular and respiratory deaths. Here the largest excesses were elsewhere.

The biggest was ICD-10 chapter 18 — symptoms, signs and abnormal clinical and laboratory findings not elsewhere classified — up more than 109% against the historical mean [s2]. Then diseases of the nervous system (95.12%), mental and behavioural disorders (85%), diseases of the skin and subcutaneous tissue (79%) and diseases of the genitourinary system (71%) [s2]. Infectious and parasitic diseases, respiratory, circulatory and neoplasm chapters also showed ratios above 1, with increments ranging from 42% down to 12% [s2].

A surge in "symptoms, signs and abnormal findings" is the coding category used when the cause of death is not established. Combined with the second finding — a 64% increase in deaths occurring at home, against 19% in hospitals and other health institutions [s2] — the picture is of people dying without a diagnosis being reached.

The authors read it that way, suggesting the pattern may point to difficulties accessing health care, gaps in risk perception among patients and caregivers, or shortcomings in clinical management for those who had prior contact with the health system [s2].

Who died

Excess mortality was significant and elevated at 13.8% for people aged 60 to 79 and 22% for those aged 80 and over [s2]. A 10% excess appeared in the 40-to-49 age group but was not statistically significant [s2].

On sex, the paper contains an internal contradiction that readers should know about. Its results section states that men showed a higher rate of mortality than women, and then reports that observed deaths were 30% above expected for women and 10% above expected for men, both significant [s2]. The abstract, discussion and conclusion all describe the event as disproportionately affecting women [s1][s2]. The numbers and three of the four narrative statements agree; the single sentence asserting a higher male rate does not. The figures — 30% for women, 10% for men — are the ones to carry forward.

Occupation, with a large caveat

Occupation was recorded for only 32.7% of records, 21,327 entries [s2]. Within that incomplete field, 37 of 672 occupations had observed-to-expected ratios above 1.25 [s2]. Grouped as indoor and outdoor work, both categories showed ratios above 1.5 — 1.66 and 1.57 — but neither was statistically significant [s2].

Individual categories that were significant included domestic workers (1.77, 95% CI 1.15 to 2.70), freelance sales representatives (2.30, 1.65 to 3.21) and merchants (1.46, 1.08 to 1.98); drivers were elevated but not significant (1.51, 0.84 to 2.73) [s2]. The highest raw ratios in the table belong to indoor occupations — housekeepers (4.33), elderly care workers (3.47), domestic workers (3.24) [s2].

With two-thirds of the field missing, these are hypothesis-generating at best. But the direction runs against the intuition that heat deaths are an outdoor-labour problem.

Limits

This is an excess-mortality analysis, not a count of heat deaths. It compares observed to expected deaths over a defined window and attributes the difference to the heat event on the grounds that nothing else explains the timing [s2]. The method cannot rule out a coincident cause, and it depends on how the expected baseline is constructed.

The paper does not carry a formal limitations section, and several supporting comparisons — including the São Paulo and Minas Gerais results, and the Covid-era mortality comparison — are described as data not shown or reserved for a separate study [s2].

Why it matters

The authors argue that some of these deaths are preventable in the short term through heat alert and response systems combined with intersectoral action plans reflecting local vulnerabilities [s2]. The cause-of-death pattern suggests where such a system would have to work: not in emergency departments treating heatstroke, but in primary care and at home, among older people with conditions that heat destabilises without ever being recorded as heat.

Sources

• [s1] Excess mortality associated with extreme heat in Rio de Janeiro, Brazil, 2023 — Revista Panamericana de Salud Pública / Pan American Journal of Public Health, published 26 July 2025. https://doi.org/10.26633/RPSP.2025.76
• [s2] Same paper, open-access full text, figures and tables — Europe PMC, PMC12296701. https://europepmc.org/article/PMC/PMC12296701